[aoirint]

Warning

This pull request was created with assistance from LLMs.

Summary

Save the model to ./base_model/animagineXL40_v4Zero.safetensors and check that same file with sha256sum. This matches the existing directory creation, Docker mount, and training model path. The SHA-256 value and download URL are unchanged.

Related Issues

• PR #111: docs: improve README identified the missing .safetensors extension in the checksum target. Thank you for pointing out this issue. The current README was independently checked and this fix was implemented separately, including the download-directory mismatch.

Notes for reviewers

The earlier PR included unexplained changes whose relationship to the fix was unclear, so it could not be trusted as submitted. This PR changes only the download destination and checksum target.

Proposed merge attribution

No additional co-authors are proposed, as requested by the maintainer. The existing commit attribution is unchanged.

AI disclosure

Codex inspected the README, made the two-line change, reviewed the diff, ran the automated checks below, and prepared this description. No model download or GPU training was performed.

Testing

Automated checks

• Passed bash -n for all four shell blocks in the README.
• Passed an offline path-consistency check covering directory creation, wget -O, the checksum target, the Docker mount, and --pretrained_model_name_or_path.
• Confirmed that the SHA-256 value and download URL are unchanged.
• Passed git diff --check and the committed-diff whitespace check.
• Markdown lint was not run locally because markdownlint-cli2 was unavailable.
• Model download, checksum verification against the actual model, GPU training, and Docker build/runtime tests were not run; this is a documentation-only path correction.

AI-assisted inspections

• Request: Align the documented model acquisition steps with the training command.
  • AI-assisted result: The downloaded file and checksum target now resolve to the same file used by the existing training mount and model argument; the change is limited to two README lines.